Skip to content

Latest commit

 

History

History
89 lines (80 loc) · 2.98 KB

commands.md

File metadata and controls

89 lines (80 loc) · 2.98 KB
Raw

Commands for Zero-Shot Classification with Stable Diffusion

Food101

python eval_prob_adaptive.py --dataset food --split test --n_trials 1 \
  --to_keep 20 10 5 1 --n_samples 20 50 100 500 \
  --prompt_path prompts/food_prompts.csv

CIFAR-10

python eval_prob_adaptive.py --dataset cifar10 --split test --n_trials 1 \
  --to_keep 5 1 --n_samples 50 500 --loss l1 \
  --prompt_path prompts/cifar10_prompts.csv

FGVC Aircraft

python eval_prob_adaptive.py --dataset aircraft --split test --n_trials 1 \
  --to_keep 20 10 5 1 --n_samples 20 50 100 500 \
  --prompt_path prompts/aircraft_prompts.csv

Oxford-IIIT Pets

python eval_prob_adaptive.py --dataset pets --split test --n_trials 1 \
  --to_keep 5 1 --n_samples 25 250 --loss l1 \
  --prompt_path prompts/pets_prompts.csv

Flowers102

python eval_prob_adaptive.py --dataset flowers --split test --n_trials 1 \
  --to_keep 20 10 5 1 --n_samples 20 50 100 500 --loss l1 \
  --prompt_path prompts/flowers_prompts.csv

STL-10

python eval_prob_adaptive.py --dataset stl10 --split test --n_trials 1 \
  --to_keep 5 1 --n_samples 100 500 --loss l1 \
  --prompt_path prompts/stl10_prompts.csv

ImageNet

python eval_prob_adaptive.py --dataset imagenet --split test --n_trials 1 \
  --to_keep 500 50 10 1 --n_samples 50 100 500 1000 \
  --prompt_path prompts/imagenet_prompts.csv

Note: for computational reasons, we evaluated on 4 images per class (4000 test images total).

ObjectNet

python eval_prob_adaptive.py --dataset objectnet --split test --n_trials 1 \
  --to_keep 25 10 5 1 --n_samples 50 100 500 1000 \
  --prompt_path prompts/objectnet_prompts.csv

Commands for Standard ImageNet Classification with DiT

If you'd like to run with 512x512 DiT, add --image-size 512 and change noise_256.pt to noise_512.pt.

ImageNet

Run 1000 separate times, with CLS from 0-999

python eval_prob_dit.py  --dataset imagenet --split test \
  --noise_path noise_256.pt --randomize_noise \
  --batch_size 32 --cls CLS --t_interval 4 --extra dit256 --save_vb

ImageNet-V2

Run 1000 separate times, with CLS from 0-999

python eval_prob_dit.py  --dataset imagenetv2 --split test \
  --noise_path noise_256.pt --randomize_noise \
  --batch_size 32 --cls CLS --t_interval 4 --extra dit256 --save_vb

ImageNet-A

ImageNet-A only has 200 classes, so run this command with CLS from IMAGENET_A_CLASSES in datasets.py.

python eval_prob_dit.py  --dataset imagenetA --split test \
  --noise_path noise_256.pt --randomize_noise \
  --batch_size 32 --cls CLS --t_interval 4 --extra dit256 --save_vb

ObjectNet

ObjectNet only has 125 classes, so run this command with CLS from OBJECTNET_CLASSES in datasets.py.

python eval_prob_dit.py  --dataset objectnet --split test \
  --noise_path noise_256.pt --randomize_noise \
  --batch_size 32 --cls CLS --t_interval 4 --extra dit256 --save_vb